This is a repository for testing various machine learning methods and applying them to chemical problems.
$ git clone https://github.com/crcollins/ml_research
$ cd ml_research
$ virtualenv env
$ . env/bin/activate
$ pip install -r requirements.txt
At this point, it is a matter of collecting the data into the data/
directory. For the mol_data
dataset, this involves running the following commands
$ mkdir data
$ cd data
$ git clone https://github.com/crcollins/mol_data
For the qm7
dataset and the qm7b
dataset run the following commands.
$ mkdir data
$ cd data
$ wget http://quantum-machine.org/data/qm7{,b}.pkl
[ ] Run multiouput with all structures
[ ] Get back working with neural nets (Caffe?)
[ ] Improve cross validation
[ ] Speed (optimize grid search)
[ ] Random search
[ ] Gradient Descent type search
[ ] Remove "bottoming out" of parameters
[ ] Add bond types to angles/dihedrals/trihedrals
[ ] forces
[ ] Restructure code
[ ] Main running code
[ ] Split up feature vectors
[ ] Split up utils functions
[ ] NN with summation for atoms
[ ] Add another condition to single_split
[ ] Parallelize feature vector creation
[ ] (Pipeline this?)
[ ] Encoding of angles